langchain-ai/deepagents · error · TypeError

Summarizer returned a {type(summary_message).__name__} summa

Error message

Summarizer returned a {type(summary_message).__name__} summary message; expected HumanMessage.

What it means

After forced compaction, the SDK's summarizer is expected to produce a `HumanMessage` as the summary message; the offload event stores this message in the checkpoint. `_forced_compaction_update` raises TypeError if the summarizer returns any other message type, since a schema-violating summary would break later compaction rounds that read the event back as their base.

Source

Thrown at libs/code/deepagents_code/offload_middleware.py:1570

            TypeError: If the summarizer's first message is not the
                `HumanMessage` the event schema declares.
        """
        summary_message = summarization._build_new_messages_with_path(
            summary, file_path
        )[0]
        if not isinstance(summary_message, HumanMessage):
            # `_build_new_messages_with_path` is annotated `list[AnyMessage]`
            # but documents (and the SDK's own call site assumes, with a type
            # suppression) that element 0 is the summary `HumanMessage`. Check
            # rather than suppress: the node turns this into a visible
            # "Compaction failed" instead of checkpointing an event whose
            # `summary_message` violates its own schema.
            msg = (
                "Summarizer returned a "
                f"{type(summary_message).__name__} summary message; expected "
                "HumanMessage."
            )
            raise TypeError(msg)
        return {
            "_summarization_event": {
                # Absolute, not relative: a second `/offload` on the same thread
                # reads this back as its base.
                "cutoff_index": state_cutoff,
                "summary_message": summary_message,
                "file_path": file_path,
            },
            "_summarization_session_id": session_id,
        }


def _create_cli_compaction_middleware(
    model: str | BaseChatModel,
    backend: BackendProtocol,
    *,
    cli_max_retries: int | None = None,
    summarization_model_spec: str | None = None,

View on GitHub (pinned to a1af029e6e)

Solutions

1. Restore/keep dcode's summarizer configuration so the summary is produced as a HumanMessage.
2. Inspect the summarizer's output parsing; add a coercion step that wraps the summary text in a HumanMessage.
3. Check for an SDK upgrade that changed summary-message type and align dcode's expectations or pin the SDK.

Example fix

// before
cutoff, summary = summarizer.summarize(messages)
return build_update(cutoff, summary)  # summary is AIMessage -> TypeError
// after
if not isinstance(summary, HumanMessage):
    summary = HumanMessage(content=summary.content)
return build_update(cutoff, summary)

summary_message = summarizer_output.summary_message
assert isinstance(summary_message, HumanMessage), f"got {type(summary_message).__name__}"

from langchain_core.messages import HumanMessage

def is_human_summary(msg: object) -> bool:
    return isinstance(msg, HumanMessage)

try:
    result = await offload(thread_id)
except TypeError as e:
    if "expected HumanMessage" in str(e):
        restore_default_summarizer_config()  # custom model/prompt broke the schema
    raise

When it happens

Trigger: A custom or upgraded summarization model/prompt whose output is parsed into an AIMessage, SystemMessage, or other message class instead of HumanMessage during `/offload` forced compaction.

Common situations: Swapping the summarizer model for one with different response parsing; overriding summarization prompts so the structured summary schema is violated; SDK version changes to SummarizationMiddleware's summary-message construction.

Related errors

• Could not parse embedded resource block. Block expected eith
• {question_type} question {question_text!r} must not define '
• Compact session start stopped by hook
• Offload server returned a non-object response.
• Offload server returned an invalid operation response.

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/e504753e4b82231b. Report an issue: GitHub.